Analyzer extractTableNamesFromStatement - returning incorrect table names

I'm using ZetaSQL in our streaming application that basically consumes from GCP audit log. As you can imagine, there are lots possibilities that a weird query will occur and flow in the audit log and eventually into our application.

I've encountered an issue specifically on the `extractTableNamesFromStatement` method of Analyzer which only occurs on specific type of query. Particularly for queries having the same table name for both target and source, like the query below.

`CREATE OR REPLACE TABLE project_test.dataset_test.test_table AS SELECT DISTINCT * FROM project_test.dataset_test.test_table`

the expected result based on my tests with various types of statements should return `[project_test.dataset_test.test_table]`

However, in the query above, the analyzer returns `[project_test, dataset_test, test_table]` instead.

here's how I used it:
```
LanguageOptions languageOptions = new LanguageOptions();
languageOptions.enableMaximumLanguageFeatures();
languageOptions.setSupportsAllStatementKinds();

AnalyzerOptions analyzerOptions = new AnalyzerOptions();
analyzerOptions.setLanguageOptions(languageOptions);
analyzerOptions.setPruneUnusedColumns(true);


Analyzer.extractTableNamesFromStatement(query, analyzerOptions)
.stream()
.flatMap(List::stream)
.collect(Collectors.toList());
```

Start at Analyzer.extractTableNamesFromStatement and reproduce the reported CREATE OR REPLACE TABLE query with the analyzer options shown. Trace why the target and source table name are split into project_test, dataset_test, and test_table; done means the result contains the single fully qualified table name project_test.dataset_test.test_table.

Written by the indexing model from the issue text.